Incorrect file reading mode and offset in Conll.py

[userofgithub1]

Hi,

I noticed in the file opening line in Conll.py the mode is incorrect it should be 'rb':

with open(path, 'rd') as f:
            doc_id = None
            doc_tokens = None

Also the calculation of the mentions positions is completely incorrect when both only reading the dataset and after linking.

The incorrect mention offsets is probably caused by these lines in class Conll.py :

begin = sum(len(t)+1 for t in doc_tokens)
dodgy_tokenisation_bs_offset = 1 if re.search('[A-Za-z],',parts[2]) else 0
position = (begin, begin + len(parts[2]) + dodgy_tokenisation_bs_offset)

Hope this is helpful and the files are edited :)
Thanks :)